Skip to main content
Glama

Offline Deployment

If you need to run this server in an offline environment (where playwright install cannot access the internet), follow these steps:

  1. On an online machine: Run the provided helper script to download the necessary browser binaries (Chromium) to a local directory (libs/browsers).

    This script automatically detects the required Playwright revision and downloads binaries for both Windows (win64) and Linux.

    uv run scripts/download_browsers.py

    This will create a libs/browsers directory containing the browser binaries (e.g., chromium-<revision>).

  2. Transfer files: Copy the entire project (or the installed package) AND the libs/browsers directory to your offline machine.

  3. Configure environment: On the offline machine, set the PLAYWRIGHT_BROWSERS_PATH environment variable to point to the libs/browsers directory before running the server.

    Windows (PowerShell):

    $env:PLAYWRIGHT_BROWSERS_PATH = "C:\path\to\mcp-fetch\libs\browsers"
    mcp-fetch

    Linux/macOS:

    export PLAYWRIGHT_BROWSERS_PATH=/path/to/mcp-fetch/libs/browsers
    mcp-fetch

    MCP Client Config (e.g., Claude/Trae): Add the environment variable to your MCP settings.

    {
      "mcpServers": {
        "fetch": {
          "command": "mcp-fetch",
          "args": [],
          "env": {
            "PLAYWRIGHT_BROWSERS_PATH": "E:\\Private\\Mcp\\fetch\\libs\\browsers",
            "MCP_FETCH_AUTO_INSTALL_PLAYWRIGHT": "0"
          }
        }
      }
    }

    Note: Setting MCP_FETCH_AUTO_INSTALL_PLAYWRIGHT=0 prevents the server from attempting to download browsers if they are missing, which fails immediately in offline mode.

Available Tools

2 tools
fetch_pageA

Fetch/Crawl a dynamic web page and convert to Markdown (supports JavaScript).

Use this tool to:

  1. Crawl/Scrape content from modern web pages (React, Vue, etc.)

  2. Get full page content after JavaScript rendering

  3. Download large page content via chunked streaming

Protocol:

  1. Start: Provide url (required) → returns transfer_id + first chunk

  2. Continue: Provide transfer_id + offset → returns next chunk

Args:

  • url: Target http(s) URL (required for phase 1)

  • to_markdown: Convert HTML to Markdown (default: True)

  • wait_selector: CSS selector to wait for before capturing content

  • Optional: headers, query, timeout_ms, max_scrolls, min/max_delay_ms, proxy/pool, user_agent, chunk_bytes

  • Cursor: transfer_id, offset (for phase 2)

Returns:

  • Chunk: chunk_text or chunk_base64, next_offset, done, truncated

  • Meta: transfer_id, status, headers, final_url, content_type, elapsed_ms

  • Size: available_bytes, total_bytes

Inputs: url, headers, query, timeout_ms, +11 more. Returns: object

http_requestA

Perform a generic HTTP request (GET, POST, PUT, DELETE, etc) for APIs or raw data.

Use this tool when:

  1. You need to call a REST API (JSON/XML)

  2. You need to use HTTP methods other than GET (POST, PUT, DELETE)

  3. You want to download a raw file without rendering (PDF, Image, etc)

Note: For GET requests to renderable web pages, prefer fetch_page which handles dynamic content and JavaScript.

Protocol:

  1. Start: Provide url (required) → returns transfer_id + first chunk

  2. Continue: Provide transfer_id + offset → returns next chunk

Inputs: url, method, headers, query, +7 more. Returns: object

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.4/5.0

    Scored across 2 tools

    Disambiguation5/5

    The two tools have clearly distinct purposes: fetch_page is for scraping dynamic web pages with JavaScript rendering and converting to Markdown, while http_request is for generic HTTP requests to APIs or raw files. The descriptions explicitly note when to use each tool, eliminating any overlap or confusion.

    Naming Consistency5/5

    Both tools follow a consistent verb_noun pattern (fetch_page and http_request) with clear, descriptive names that match their functions. There are no deviations or mixed conventions in the naming style.

    Tool Count3/5

    With only 2 tools, the server feels thin for a general-purpose HTTP fetching domain, as it might lack coverage for specialized scenarios like WebSocket handling or advanced caching. However, the tools cover core use cases (dynamic page scraping and generic HTTP requests), making it borderline but functional.

    Completeness4/5

    The tool set covers essential HTTP operations for web scraping and API interactions, with clear guidance on when to use each tool. Minor gaps exist, such as no dedicated tools for WebSocket connections or batch request handling, but agents can likely work around these using the provided tools.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      D
      maintenance
      A minimal MCP server for homelab environments, providing demo tools like ping and echo over Streamable HTTP for testing client-server integration.
      -
    • F
      license
      A
      quality
      D
      maintenance
      An MCP server that enables LLMs to make arbitrary HTTP requests (GET, POST, PUT, DELETE, etc.) with custom headers, bodies, and cookies, supporting JSON and error handling.
      1
      2
      -
    • A
      license
      Not graded
      quality
      D
      maintenance
      A stateless MCP server that provides a configurable HTTP client tool for LLM agents, supporting custom methods, headers, bodies, timeouts, redirects, response size limits, DNS overrides, and TLS verification.
      MIT
    • F
      license
      Not graded
      quality
      D
      maintenance
      A minimal MCP server providing tools for time, calculations, sandboxed file operations, and safe URL fetching over stateless HTTP.
      -